Vision AI Cuts Pallet Repacking by 70% for Drugstore Retailer

Context

A major retail chain specialising in drugstore products, cosmetics, health and wellness, household items and beauty products operates three high-throughput distribution centres that were already running near capacity. The objective was to improve the pallet packing process across these centres without disrupting existing operations: increase picking speed while maintaining accuracy, reduce packaging errors, and address spacing inefficiencies on pallets to lift overall throughput. The solution needed to integrate with the existing monorail robot delivery system and human packers, preserve safety standards, and deliver measurable gains in operational efficiency, sustainability and cost control.

Challenges

Three interrelated constraints limited throughput. First, package detection relied on an industrial scale on each robot that required a tare before and after every placement; this weight-based check cost 3–5 seconds per item at maximum OEE and became a persistent bottleneck. Second, the single-track monorail robot system created traffic jams during peak promotions and high-volume product runs: robots queued behind one another with no overtaking possible, and congestion amplified delays across the network. Third, packers lacked aisle-level visibility into ongoing picks, so pallet loading became a manual “Tetris” exercise; despite extra handling, average pallet utilisation stalled at around 75%, increasing repacking and reducing effective vehicle capacity. All three issues needed fixes that increased speed and accuracy without requiring major infrastructure changes.

Implementation

The retailer replaced weight-based detection with a Vision AI solution that uses a top-mounted 3D camera and a deep-learning engine to recognise packages by their dimensions and visual features. Detection time dropped from 3–5 seconds to about 1 second per item, while object-level verification and safety checks preserved operational standards. A projector was mounted and calibrated over the packing station to indicate optimal placement for each package; an AI model consumed the preloaded picklist and calculated ideal locations in real time, updating the projected pallet layout so packers could place items without manual rearrangement. The faster, camera-based detection also enabled multiple packers to safely work on the same pallet simultaneously, reducing single-file dependencies and easing pressure on the monorail system. Together, these measures reduced handling steps, eliminated the need for repeated taring, and improved coordination between robots and packers with software-driven placement guidance rather than manual guesswork.

Results

The Vision AI implementation delivered immediate, measurable gains: overall picking and packing efficiency improved by over 30% and pallet repacking fell by 70%, allowing significantly more pallets to be completed per shift. Average pallet utilisation increased from roughly 75% to around 85% as projector-guided placement removed the need for time-consuming load rearrangement. The combination of faster detection and multi-packer capability reduced robot congestion by at least 5% as queuing delays declined. End-to-end benefits included approximately 5% lower operating costs and up to 10% higher productivity, while fewer pallets per vehicle cut CO₂ emissions and improved sustainability metrics. By addressing detection, traffic and spacing constraints with a low-disruption Vision AI retrofit, the retail chain increased throughput and accuracy across its three distribution centres without major hardware overhauls.